GazePlotter
Documentation Chapters

Scanpath Similarity

Scanpath similarity metrics quantify the spatial and temporal similarity between the gaze trajectories of different participants. In GazePlotter, this comparison is computed pairwise, representing the alignment of scanpaths across the entire group.

Output Shape and Projections Translation

All scanpath similarity metrics naturally output a participant-pair-matrix (an M×M grid representing pairwise similarity scores, where M is the number of active participants). Because the output is inherently group-level, no further dimensional projection is supported.

1. Matrix Passthrough (participant-pair-matrix)

You can pass the matrix through directly to analyze the similarity relationships between all pairs of participants in the selected group:

  • Identity (identity-participant-pair-matrix): Outputs the M×M similarity matrix.

Visualizer Compatibility: Passing the raw matrix through allows you to select scanpath similarity in the Scanpath Similarity visualization (non-windowed). It cannot be projected to other shapes and cannot be selected in other plots.

Invariants

  • Measurement class: Group-level. The result is inherently a property of a pair of participants, where both the row and column axes are the participants themselves. There is no per-participant value to combine across the group — the matrix is already the across-participant result, so no further aggregation applies.
  • Windowing: Forbidden (supportsWindowing: false). Sliding-window projections are not supported. However, cropping by a Time of Interest (using timeStart and timeEnd bounds in the Settings Pane) is supported; fixations are encoded into the scanpath string only if their onset falls within the range.

Scanpath String Encoding

To compare scanpaths, GazePlotter converts each participant’s fixation sequence into an AOI-letter string:

  • Each active AOI is mapped to a unique character (e.g., AOI 0 -> ‘A’, AOI 1 -> ‘B’).
  • Off-AOI fixations that do not land within any active boundary are encoded using a special hash character (#).

Metric Recipes

1. Participant Pair Similarity (participantPairSimilarity)

Computes the similarity score for every pair of participant scanpaths. The resulting matrix is symmetric (similarity from participant i to j equals j to i) and has a diagonal of 1.0 (every participant is identical to themselves).

Parameters

  1. Similarity Method (method):

    • Levenshtein (levenshtein): Calculates the minimum edit distance (insertions, deletions, and substitutions) required to transform one scanpath string into the other. The score is normalized by the maximum length of the two scanpaths:

      Similarity = 1 - (Edit Distance / max(Length_1, Length_2))

    • Needleman-Wunsch (needlemanWunsch): A dynamic programming algorithm for global sequence alignment. It scores alignments using standard parameter weights (Match = +1, Mismatch = -1, Gap Penalty = -1) and normalizes the final alignment score to fit a 0–1 scale.

  2. Collapsed Scanpaths (collapsed):

    • false (default): Consecutive fixations inside the same AOI are preserved. A participant who fixates AOI A three times before moving to B is encoded as “AAAB”.
    • true: Consecutive identical AOI entries are collapsed into a single character (“AAAB” $\rightarrow$ “AB”). This removes duration-based variance, isolating the analysis to focus purely on the structural order of visited regions.

Visualizing the Similarity Matrix

The resulting similarity scores can be explored in two ways under the Scanpath Similarity visualization:

  1. Heatmap Matrix: Displays the raw N×N grid, colored by similarity strength, to identify clusters of participants with similar viewing strategies.
  2. ScanGraph Network: Represents participants as nodes in a graph. An edge is drawn between two participants only if their similarity score exceeds a user-defined threshold, helping visualize communities of shared attention.